• First empirical comparison of explanation strategies in smart environments: We conduct an empirical study (N = 159) to systematically compare three explanation: no explanation, static explanations (fixed and identical across users and contexts), and context-adapted explanations (dynamically adjusting content based on user preferences and situational factors). To the best of our knowledge, this is the first empirical study in the rule-based smart environments domain to directly contrast static and context-adapted explanations in terms of their impact on task performance, system understanding, usability, and perceived explanation quality. • Explanations boost performance and understanding: We found that providing explanations, whether static or context-adapted, significantly improves user task performance and comprehension of system behavior compared to offering no explanations. • Context-adaptivity is not always superior: We revealed that context-adapted explanations are not universally better than static ones, with their benefits depending on factors such as task complexity and user preferences. The increasing complexity of interactive smart environments presents a significant engineering challenge: their automated, context-aware decisions are often opaque, undermining system usability. While explainability is becoming known as a promising remedy, systematically evaluating different explanation strategies for these systems remains an open problem. This paper presents a rigorous empirical evaluation to address this gap. We conducted a controlled experiment (N = 159) to compare three approaches: no explanation, static explanations, and context-adapted explanations. Our results quantify the significant benefits of explainability on task performance and user understanding of system behavior. Furthermore, we identify a key engineering tradeoff: context-adapted explanations are not universally superior to simpler static implementations, suggesting that a one-size-fits-all approach is suboptimal in this domain. To support our evaluation, we developed a web-based testbed simulating a smart home environment with light gamification elements, enabling reliable human-grounded assessment. Based on our findings, we offer concrete insights and recommendations to guide the design of explainable interactive systems. Our study underscores the importance of tailoring explanations to user needs and contextual factors, contributing to more transparent and user-friendly smart environments.
Sadeghi et al. (Sun,) studied this question.